Presentation on theme: "Ioe.ac.uk Uncertain Knowledge: Exploring quantitative data from a qualitative perspective Gemma Moss Institute of Education, University of London CHEER,"— Presentation transcript:
ioe.ac.uk Uncertain Knowledge: Exploring quantitative data from a qualitative perspective Gemma Moss Institute of Education, University of London CHEER, Oct 2013
Session overview: Can qualitative research traditions interact with quantitative traditions in ways that work for the common good? looking at quantitative practice with an ethnographic lens The telling case: reading engagement in PISA Unpacking the hypotheses that structure statistical design
The stand-off between quant and qual: Positivist - Scientific External reality - need to collect 'facts' Methods of natural sciences Use of statistics (quantitative) Experiments, surveys Interpretivist - Social Scientific Subjective, constructed reality Relative truths Need to explore, explain and understand reality Qualitative http://www.slideshare.net/clairetupling/positivist-interpretivist
Reading the certainty of numbers: Eighteenth-century European conceptions of government articulated a notion of statistics, or science of state, in which the operation of government was to be made possible by the accumulation and tabulation of facts about the domain to be governed. … a huge labour of inscription which renders reality into a calculable form.... The 'representation' of that which is to be governed is an active technical process. Government has inaugurated a huge labour of enquiry to transform events and phenomena into information: births, illnesses and deaths, marriages and divorces, levels of income and types of diet, forms of employment and want of employment. By means of inscription, reality is made stable, mobile, comparable, combinable.... Information in this sense is not the outcome of a neutral recording function. It is itself a way of acting upon the real (Rose and Miller, 1992)
Report of the Birmingham Statistical Society on the State of Education in Birmingham, April 1840
On the trail of numbers: The Literacy Attainment Data and Discourse Project (ESRC fellowship, 2011-13) To explore: the design and use of literacy attainment data in three different historical periods: the 1860s; 1950s; and 2000s. how these data arise in their institutional and socio-historical contexts, and are analysed and understood as part of public and professional discourse The datasets include: elementary school examination data collected by government as part of “payment by results”; eleven plus data, collected by Local Authorities and used to determine entry to grammar school; and PISA data, collected by the OECD for purposes of international comparison
Getting to here from there – Or the Autobiography of the problem.... Mixed methods research: testing quants against quals in a study of library borrowing records (Moss and McDonald, 2004) Using quant measures to hold policy to account (Moss et al, 2007) Integration or interrogation in mixed methods research? (Brannen and Moss, 2012) Measures and proxies (Jerrim and Micklewright, 2012)
Understanding the genre of quantitative presentation: What are the requirements of its spoken or written forms? Quantitative methods as social practice - formulate a question with a numerical answer - take into account: the uncertainty in the numbers - the potential unreliability of the findings - determine the strength of the answer through the aptness of the measures and the choice of models - When the numbers don’t work out – search for alternative formulations and explanations...
“Objectivity” or uncertainty in statistical discourse? The t-test assesses whether the means of two groups are statistically different from each other. … Once you compute the t-value you have to look it up in a table of significance to test whether the ratio is large enough to say that the difference between the groups is not likely to “The Research Methods Knowledge Base: The T-Test have been a chance finding. To test the significance, you need to set a risk level (called the alpha level). In most social research, the "rule of thumb" is to set the alpha level at.05. This means that five times out of a hundred you would find a statistically significant difference between the means even if there was none (i.e., by "chance").” http://www.socialresearchmethods.net/kb/stat_t.phpalpha level
BBC, 2004 “My argument today is as follows: - that the Pisa study confirms that teachers and pupils in our country are professional and hard working; on average we are in the top quartile of performers - therefore our first challenge is to think how we move from having a system that is improving and in significant respects good, to one that is truly great; this will take investment and reform.” David Milliband, 2004
Pisa country rankings by % of students at Level 3 and above. Knowledge and Skills for Life, Executive Summary, 2000 Reading the stats: where are we?
PISA and the technicalities of testing: querying the data “The article comments on the restricted nature of the data modelling and analysis, and the resulting interpretations. It points to certain features of the results that raise questions about the adequacy of the data and it stresses the failure to introduce a longitudinal component.” “In PISA several questions are grouped in that they all relate to the same problem. For example, one problem described a pattern of trees planted as a set of squares of different sizes, and associated with this problem there were three separate questions. It is doubtful whether the independence (iid) assumption would hold for those composite questions and for simplicity we have selected 15 items each of which is a response to a different problem and dichotomised into correct/incorrect, treating part-correct answers as correct. Goldstein, H (2004)Assessment in Education: principles, policy & practice, 11:3
PISA as a policy tool for governance: unpicking the policy discourse ““PISA offers a new approach to considering school outcomes, using as its evidence base the experiences of students across the world, rather than in the specific cultural context of a single country. The international context allows policy-makers to question assumptions about the quality of their own country’s educational outcomes.” (OECD 2001, 27) De-contextualisation, commensurability and policy orientation have been the key ingredients contributing to PISA’s success. Grek (2009) “However, empirical evidence suggests that PISA is in fact not the same everywhere. If it is a major influencer of policy in Australia, it is also not central to policy in the US. If it causes ‘shock’ in Germany, setting off intense policy activity, it is merely a source of reassurance in the UK that brings about no change in policy or practice. Gorur (2008)
The role of numbers in public discourse: “Rose (1999) observes that single numbers in policy accounts do political work in hiding the technologies that have gone into their construction. “Numbers are resorted to in order to settle or diminish conflicts in a contested space of weak authority. And the ‘power of the single figure’ is here a rhetorical technique for ‘black boxing’—that is to say, rendering invisible and hence incontestable—the complex array of judgments and decisions that go into a measurement, a scale, a number. The apparent facticity of the figure obscures the complex technical work that is required to produce objectivity.” (Rose 1999, p. 208)... part of the educational research agenda needs to acknowledge that data in policy and research are made, fabricated—not in the sense of falsified, but in the sense of constructed, put together; these matters need to be a focus of our research as well.” (Lingard, 2010, AER)
“Rose (1999) observes that single numbers in policy accounts do political work in hiding the technologies that have gone into their construction. “Numbers are resorted to in order to settle or diminish conflicts in a contested space of weak authority. And the ‘power of the single figure’ is here a rhetorical technique for ‘black boxing’—that is to say, rendering invisible and hence incontestable—the complex array of judgments and decisions that go into a measurement, a scale, a number. The apparent facticity of the figure obscures the complex technical work that is required to produce objectivity.” (Rose 1999, p. 208)... part of the educational research agenda needs to acknowledge that data in policy and research are made, fabricated—not in the sense of falsified, but in the sense of constructed, put together; these matters need to be a focus of our research as well.” (Lingard, 2010, AER) Or deconstructing statistical data....
Policy document or domain-specific display? Pisa country rankings by % of students at Level 3 and above. Knowledge and Skills for Life, Executive Summary, 2000
Chart 2.1 Average scores and confidence intervals for Canadian provinces and other countries: Reading. PISA data, 2006. http://www.pisa.gc.ca/ eng/measuring-up- 2006.shtml Note: The confidence interval represents the range within which the score for the population is likely to fall 95% of the time or 19 times out of 20. Differences in average scores between two jurisdictions are not statistically significant when the confidence interval for each average score overlaps.
Differences in average scores between two jurisdictions are not statistically significant when the confidence interval for each average score overlaps
What’s in the data? a value relevance framework..... “The value relevance framework affects the process of explanation.. [and] frames the selection of explanatory factors. The task of explaining differences in educational achievement or occupational mobility does not amount to discovering the unique or exhaustive set of determinants of these phenomena ….. we necessarily select out from this infinite field those causes that are relevant to the value framework within which the research is being carried out.” Hammersley, 2009 Putting the data to work....
Understanding PISA data: modelling correlations and causes The ingredients that structure the analysis: - Test scores, sorted into 5 levels - School level questionnaire (sent to Head), identifying system characteristics - Student questionnaire, identifying learner characteristics Optional: –Parent questionnaire identifying family social, economic and cultural capital –Computer familiarity –Educational career
This questionnaire asks for information including : The structure and organisation of the school; The student and teacher body; The school’s resources; The school’s instruction, curriculum and assessment; The school climate; The school policies and practices; The characteristics of the headteacher/principal or designate This information helps illustrate the similarities and differences between groups of schools in order to better establish the context for students’ test results. For example, the information provided may help to establish what effect the availability of resources may have on student achievement – both within and between countries. (NFER, 2009) School Questionnaire:
Your answers will be kept confidential. They will be combined with answers from other schools to calculate totals and averages in which no one school can be identified. (NFER, 2009) School Questionnaire: Which hypothesis? Why this relationship of parents to school?
In this booklet you will find questions about: You Your family and your home Your reading activities Learning time Classroom and school climate Your English lessons Libraries Your strategies in reading and understanding text (NFER, 2009) Pupil questionnaire:
Learning to Learn Are students who enjoy reading better readers? What kinds of reading are associated with being a good reader? Do boys and girls have different reading habits? What learning strategies help students perform better? Reading for engagement Reading for enjoyment, by gender and background Which correlations? What do they help explain? Why enjoyment and engagement? Where do these terms come from? PISA 2009 at a glance - Reading
Part of public as well as academic discourse making visible differences in who engages in what kinds of reading “Reading often” and “reading well” are frequently linked The idea that boys are particularly reluctant to read circulates informally long before it gets taken up and imported into systematic enquiry Reading engagement - taking the long view:
Reading engagement and attainment – factoring in gender....
The Excitement; or A book to induce young people to read 1839 edition, frontispiece
The Excitement; or A book to induce young people to read “While we are often told by parents... that such a boy is a fine clever boy, but he has, unhappily, no taste for reading; may not this often arise from not pointing out to such a young person the pleasure to be derived from reading, by supplying him... with materials suited to his taste, and in this way forming in him the habit of seeking enjoyment from this source? The object of this volume... is to present such materials by furnishing the youthful reader with an account of those striking appearances of nature, and signal preservations, the description of which is generally listened to, by boys in particular, with the greatest attention” Editor’s Preface, 1831, p vi
Questions on: Resources – how many books Time spent reading for enjoyment Attitude towards reading Diversity in reading/ reading interests Time spent reading online for functional purposes Exploring reading engagement: going inside the data box
Q22. How many books are there in your home? There are usually about 40 books per metre of shelving. Do not include magazines, newspapers, or your school books. (Please tick only one box.) 0-10 books 1 11-25 books 2 26-100 books 3 101-200 books 4 201-500 books 5 More than 500 books 6 Looking for correlations: Resources Scaling the categories....? What are the significant boundaries?
Time spent Modelling reading engagement: Q23 About how much time do you usually spend reading for enjoyment? (Please tick only one box) I do not read for enjoyment 1___ 30 minutes or less a day 2___ More than 30 minutes to less than 60 minutes a day 3___ 1 to 2 hours a day 4___ More than 2 hours a day 5___ Correlation or cause? How to decide?
Modelling reading engagement: Counting motivation to read?
PISA found that boys and girls have different types of reading interests, which is fairly consistent across countries. Boys’ interest in a wide range of materials including non-fiction, newspapers and comics, but their much lower interest in reading fiction books, suggests that the choices of reading materials may influence the success of any programme to engage boys more in reading. Different strategies may be appropriate for boys and for girls, who tend to have different reading interests, with girls particularly interested in books, especially fiction, and boys more interested in other forms such as newspapers and comics. Messages from PISA 2000 Categorising the data.........
PISA also found that boys and girls have different types of reading interests, which is fairly consistent across countries. Boys’ interest in a wide range of materials including non-fiction, newspapers and comics, but their much lower interest in reading fiction books, suggests that the choices of reading materials may influence the success of any programme to engage boys more in reading. Different strategies may be appropriate for boys and for girls, who tend to have different reading interests, with girls particularly interested in books, especially fiction, and boys more interested in other forms such as newspapers and comics. Extrapolating from the data...... Expressions of confidence....
The gender gap in enjoyment of reading helps to explain why girls continue to outperform boys significantly in reading. It is also worrying that the impact of socio-economic background on reading for enjoyment, which had been relatively weak in 2000, is growing stronger. These trends highlight the particular urgency of finding ways to engage boys from disadvantaged backgrounds in reading for pleasure. Or asking statistics to go critical Extrapolating from the data
Modelling reading engagement in PISA: What are the underlying hypotheses?
Ho wever, in his evidence to the Commission, Phil Jarrett nuanced this statement: “Schools need to value and teach a wider range of texts than currently. We know that boys tend to read different kinds of texts from girls – non- fiction, autobiographies, newspapers and so on – yet the English curriculum largely values certain kinds of narrative fiction texts, I think. I think for boys it often seems that what they read outside school does not matter; it does not count in relation to the classroom. We need to bring those resources much more into the classroom.” This view also came through from schools that took part in the survey: “I think [the gender gap] is due to boys turning off from reading at secondary school and curriculum texts not lending themselves to boys’ interests. I also feel that libraries are often too heavily stocked with fiction books.” The All Party Parliamentary Literacy Group Boys' Reading Commission Why numbers should be part of the discourse as a check on belief
Reading comic books is generally associated with a low level of reading performance. This could well be because weaker readers find comic books more accessible. In some of the countries that show small gender differences in enjoyment of reading, both boys and girls are relatively unlikely to report that they enjoy reading. In Japan, for example, only 54% of boys and 58% of girls reported that they enjoy reading. In some countries, the narrow gender gap reflects the opposite. The fact that two in three boys, on average in OECD countries, reported that they read newspapers for pleasure, compared to only one in five who said they read fiction for enjoyment, shows that there could be far more potential for strengthening boys’ reading skills by encouraging other types of reading in addition to literature. Challenging what stays in, and what goes out of the model?
Questions for discussion: Is distinguishing between statistical knowledge and policy discourse useful? Do qualitative traditions have a useful role to play in examining the claims made using quantitative techniques? Do current criticisms of quantitative traditions run the risk of exaggerating their power by viewing them through a policy lens? How much should we invest in knowing about numbers? How else might the politics of quantitative data in public discourse be addressed?
Utterly wrong! Flawed! Academics deride league tables that guide Michael Gove's reforms “Professor Svend Kreiner, a statistician from the University of Copenhagen in Denmark, said the Pisa model is fundamentally flawed. In a paper published this summer, he challenges Pisa's reliability and shows how results fluctuate significantly according to which test questions are used. He also reveals how, in the 2006 reading rankings, Canada could have been positioned anywhere between second and 25th, Japan between eighth and 40th and the UK between 14th and 30th. Mr Gove has said that Britain between 2000 and 2009 "plummeted in the world rankings from 4th to 16th for science, 7th to 25th for literacy and 8th to 28th for maths". Pisa's 2012 results are published later this year.” Independent, Friday 19 July 2013